Arginine and the Hofmeister Series: The Role of Ion–Ion
Interactions in Protein Aggregation Suppression
The MIT Faculty has made this article openly available. Please share
how this access benefits you. Your story matters.
Citation
Schneider, Curtiss P., Diwakar Shukla, and Bernhardt L. Trout.
Arginine and the Hofmeister Series: The Role of Ion–Ion
Interactions in Protein Aggregation Suppression. The Journal of
Physical Chemistry B 115, no. 22 (June 9, 2011): 7447-7458.
As Published
http://dx.doi.org/10.1021/jp111920y
Publisher
American Chemical Society (ACS)
Version
Author's final manuscript
Accessed
Wed May 25 20:31:51 EDT 2016
Citable Link
http://hdl.handle.net/1721.1/80695
Terms of Use
Article is made available in accordance with the publisher's policy
and may be subject to US copyright law. Please refer to the
publisher's site for terms of use.
Detailed Terms
NIH Public Access
Author Manuscript
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.
NIH-PA Author Manuscript
Published in final edited form as:
J Phys Chem B. 2011 June 9; 115(22): 7447–7458. doi:10.1021/jp111920y.
Arginine and the Hofmeister Series: The Role of Ion-Ion
Interactions in Protein Aggregation Suppression
Curtiss P. Schneider‡, Diwakar Shukla‡, and Bernhardt L. Trout*
Department of Chemical Engineering, Massachusetts Institute of Technology, 77 Massachusetts
Avenue, Cambridge Massachusetts 02139
Abstract
NIH-PA Author Manuscript
L-arginine hydrochloride is a very important aggregation suppressor for which there has been
much attention given regarding elucidating its mechanism of action. Little consideration, however,
has been given toward other salt forms besides chloride, even though the counterion likely imparts
a large influence per the Hofmeister Series. Here, we report an in depth analysis of the role the
counterion plays in the aggregation suppression behavior of arginine. Consistent with the
empirical Hofmeister series, we found that the aggregation suppression ability of other arginine
salt forms on a model protein (α-chymotrypsinogen) follows the order:
H2PO4−>SO42−>Citrate2−>Acetate−≈F−≈Cl−>Br−>I−≈SCN−. Mechanistically, preferential
interaction and osmotic virial coefficient measurements, in addition to molecular dynamics
simulations, indicate that attractive ion-ion interactions, particularly attractive interactions
between arginine molecules, play a dominate role in the observed behavior. Furthermore, it
appears that dihydrogen phosphate, sulfate, and citrate have strong attractive interactions with the
guanidinium group of arginine, which seems to contribute to the superior aggregation suppression
ability of those salt forms by bridging together multiple arginine molecules into clusters. These
results not only further our understanding of how arginine influences protein stability, they also
help to elucidate the mechanism behind the Hofmeister Series. This should help to improve
biopharmaceutical stabilization through the use of other arginine salts and possibly, the
development of novel excipients.
Keywords
NIH-PA Author Manuscript
Arginine; Hofmeister Series; Preferential Interaction Coefficient; Osmotic Virial Coefficients;
Vapor Pressure Osmometry; Protein Aggregation
Introduction
Protein aggregation is a major problem that pharmaceutical and biotech companies face
when producing protein based therapeutics.1,2 Not only does aggregation during production
reduce yield, it also places a huge burden on downstream purification processes.3 More
importantly, aggregation during formulation and storage can have serious consequences
because aggregates can elicit an immune response and other adverse side effects if
administered to a patient.4,5 Over the past 25 years, the number of protein based therapeutics
*
Corresponding author. trout@mit.edu, Phone: (617) 258-5021, Fax: (617) 253-2272.
‡These authors contributed equally to this work.
Supporting Information Available
A complete and detailed description of each experimental methodology can be found in the supporting information and this
information is available free of charge via the Internet at http://pubs.acs.org.
Schneider et al.
Page 2
NIH-PA Author Manuscript
on the market or in development has grown exponentially, thus sparking an interest in
developing improved methods for preventing aggregation.6,7 For decades, the method and
materials most often used for inhibiting the aggregation of proteins in solution has
essentially remained unchanged; which is to add aggregation suppressing solution additives
(often called cosolutes, cosolvents, and excipients) to the protein solution during production,
purification, and formulation.2,8,9 Though most formulation recipes include a pH buffer, a
nonionic surfactant, and maybe some sort of salt, the key excipient for preventing
aggregation is usually some type of sugar or polyol (e.g. sucrose, trehalose, glycerol,
sorbitol, etc.), which protects the native structure of the protein from unfolding, thus
reducing the number of aggregation prone species in solution.10,11 However, other excipient
types have been used and/or are gaining much more attention (e.g. amino acids, polymers,
proteins, etc.). Furthermore, there is a desire to discover or create better performing
excipients because for many protein therapeutics, a stable formulation cannot be created and
the product must be lyophilized and reconstituted prior to injection, which for the most part
is undesirable and, in some cases, a difficult and costly challenge.2,7,11
NIH-PA Author Manuscript
One excipient that has gained much attention lately is arginine hydrochloride (ArgHCl).12
Its ability to greatly improve the refolding yield of proteins was discovered, nearly by
accident, two decades ago, and since then it has been the topic of many research articles,
mainly due to its effectiveness and unique behavior in regards to suppressing
aggregation.13–22 Arakawa and coworkers have written many research articles and reviews
regarding the abilities and applications of ArgHCl, including its use during protein
production and purification and as a solution additive for other macromolecules.16,23–30
Their collective work provides useful insight into the mechanism of action and highlights all
of the observed behavior gathered thus far for ArgHCl. Those studied have revealed much
about the arginine mechanism but a complete mechanistic understanding has yet to be
determined.
NIH-PA Author Manuscript
What is known about ArgHCl is that it seems to inhibit protein-protein interactions with
little influence on the conformational stability of the protein, thus slowing association rather
than stabilizing the native structure.14,15,31 Some researchers have attempted to explain this
behavior, but no one theory has been generally accepted because most theories cannot
explain all of the observed behavior, not to mention that there is often a lack of experimental
evidence confirming the proposed mechanism of action.12 A theory first proposed in our
group, suggests that arginine is neutral in regards to preferential interactions, leading to little
influence on the folding equilibrium but eliciting an entropic penalty for when two protein
molecules aggregate due to the exclusion of arginine molecules from the gap formed
between two associating protein molecules.14 The preferential interaction coefficient values
we measured later on seemed to, for the most part, support this theory.15 However, at high
concentrations, the preferential interaction of arginine goes from being neutral or slightly
attracted to being highly excluded, and thus, we have to admit that this model only describes
a simple theoretical molecule, referred to as a “neutral crowder”, and cannot explain all of
the complex behavior ArgHCl exhibits, though the “neutral crowder” effect may play a large
role.12,22,32
Moreover, little attention has been devoted towards other salt forms besides chloride. The
chloride form has been studied so exclusively that ArgHCl is often referred to simply as
arginine.12 The Hofmeister Series,
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.
Schneider et al.
Page 3
NIH-PA Author Manuscript
NIH-PA Author Manuscript
first published in 1888, is an empirical ranking of how various ions influence protein
solubility (i.e. ions to the left have a greater ability for precipitating proteins), which in turn
is related to the thermodynamic stability of the native state.33 The series has been studied
extensively since the time of Hofmeister and serves as the foundation for selecting salts that
will influence protein solubility, crystallization, denaturation, and aggregation, though the
mechanism for this behavior is not well understood.34–37 It has been clearly demonstrated
that the choice of anion has a more pronounced effect on these properties than the choice of
cation. Furthermore, chloride is often placed in the middle of the series, thus other arginine
salt forms are likely to exhibit a drastically different behavior, both positively and
negatively, in regards to protein stability.37,38 To the best of the authors’ knowledge, the
only other form of arginine studied in this context has been ArgH(SO4)1/2, though the extent
of those studies is limited.39,40 Recently though, the combination of arginine hydrochloride
with sodium glutamate was studied by two different research groups and has received much
attention due to a seemingly synergetic effect when the two solutes are combined in an
equimolar ratio.7,41,42 Moreover, in a recently published review, Lange and Rudolph12
(Rudolph being one of the researchers who discovered the arginine effect) postulated, based
on existing data, about how other arginine salt forms would behave as excipients and also
commented on the refolding behavior of the sulfate and phosphate salt forms, stating that
those two forms of arginine perform poorly as refolding excipients even though sulfate and
phosphate often provide conformational stabilization. However, no citations or experimental
results were provided to support those claims. Even though sulfate and phosphate may
reduce the refolding abilities of arginine, we have found that they enhance its aggregation
suppression ability when our model protein is aggregated under accelerated conditions.
Moreover, switching to a thiocyanate salt induces rapid aggregation under similar conditions
rather than enhancing aggregation suppression, opposite of what was hypothesized.
However, testing on other proteins is required before one could generalize these
observations.
When attempting to explain the arginine mechanism, many researchers start by addressing
the effect of the guanidinium moiety.40,43 Indeed, the guanidinium functional group must
play an important role, because not only is arginine the only amino acid with a guanidinium
moiety, no other amino acid exhibits aggregation suppression characteristics quite like
arginine, though other amino acids, particularly proline, have been used as stabilizers.11,44,45
However, without a clear understanding at the molecular level, not only of the interaction of
arginine with protein molecules, but also of how arginine behaves in solution, the arginine
mechanism will remain a mystery.
NIH-PA Author Manuscript
Recently though, two key breakthroughs, in relation to the solution behavior of guanidinium
and arginine, may help to resolve this perplexing question. The first comes from a series of
articles published by Mason and coworkers, in which they studied the solution behavior of
various guanidinium salts.46–51 They initially aimed to explain the denaturing ability of
guanidinium but ultimately revealed that guanidinium can form a strong attractive
interaction with various anions of similar size that also have hydrogen bond accepting
capabilities (e.g. sulfate, carbonate, etc.), which likely explains why some guanidinium salts
(e.g. guanidinium sulfate) stabilize proteins rather than denature them because the two ions
form clusters in solution rather than bind to the protein surface.50 The hydrogen bonding
interactions and cluster formations they demonstrated reveal a lot about the behavior of
guanidinium. However, some of this behavior was revealed long before the work of Mason
and coworkers. Bonner, using osmotic coefficient data and Raman spectra, gave valuable
insight into the hydrogen bond interactions between guanidinium and halide ions over three
decades ago.52,53 Bonner showed that guanidinium and fluoride form a strong hydrogen
bond, possibly even a partial H-F bond, in solution and this interaction is weaker for other
halides in the order of Cl- > Br- > I-, due to ion size and charge density. At the time, these
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.
Schneider et al.
Page 4
NIH-PA Author Manuscript
results were used to speculate on the hydrogen bond formation between guanidinium and
proteins in an attempt to explain the denaturing effect of guanidinium. More importantly,
Mason and coworkers also revealed that the structure of the poorly hydrated guanidinium
molecule allows the positively charged ions to associate, with the planar molecules stacked
on top of each other.47 These results contradict long held beliefs about how certain cosolutes
influence the structure of water (solutes have long been labeled either a chaotrope or a
kosmotrope because of this water structure viewpoint) and how the perturbed water structure
influences protein stability.37 Their results are compelling and have strong implications for
explaining the arginine mechanism, not only with how arginine will interact with various
counterions, but how arginine interacts with itself and glutamate (when mixed with that
amino acid), given that guanidinium should have a strong attractive interaction with the
carboxylate moiety found on both molecules.
NIH-PA Author Manuscript
This self-association of arginine, the second breakthrough mentioned, has previously been
revealed and has been studied extensively in our lab through molecular dynamics (MD)
simulations.17,22,32,54 These simulations show that the guanidinium functional group does
indeed interact with the surface of a protein molecule in a fashion similar to free
guanidinium; however, this functional group also forms attractive interactions with other
arginine molecules. This self-association of arginine molecules likely prevents arginine from
binding too strongly to the surface of protein molecules, preventing it from denaturing the
protein but at the same time preventing protein-protein interactions. This self-association
also explains the unique trend observed for preferential interaction coefficient measurements
(i.e. slightly bound at low concentrations but highly excluded at high concentrations) and
explains why the arginine effect diminishes at high concentrations.15 Therefore, we were
motivated to explore other arginine salt forms, not only with the aim of discovering a better
performing excipient, but to further our understanding of this mechanism and how
guanidinium-anion interactions may contribute to the observed behavior.
NIH-PA Author Manuscript
In this work, we studied the influence a variety of arginine salt forms have on the heat
induced aggregation of α-chymotrypsinogen A. We show that dihydrogen phosphate,
sulfate, and citrate enhance aggregation suppression over chloride, while bromide, iodide,
and thiocyanate diminish it. This trend is in accordance with the influence these cosolutes
have on the conformational stability of the protein, as measured by differential scanning
calorimetry (DSC), indicating that changes in the folding equilibrium is an important factor
in the aggregations suppression mechanism. However, perturbing the structure of the protein
cannot explain everything, since the acetate, fluoride, and chloride salt forms have no
influence on the midpoint melting temperature. Trends in the preferential interaction
coefficient, as determined by vapor pressure osmometry (VPO) and MD simulations,
support the observed behavior and show a key difference between the anions that enhance
aggregation suppression and the anions that diminish it. Osmotic second virial coefficient
measurements, as determined by VPO and MD simulations, give a clear indication for the
observed behavior and support previous reports of ion-ion interactions.48,49,55 Dihydrogen
phosphate, sulfate, and citrate seem to interact favorably with the guanidinium moiety, with
phosphate interacting quite strongly, which reduces the solubility of arginine considerably.
These ion-ion interactions seem to limit interactions with protein molecules, while the
anions that have a weaker interaction with guanidinium allow arginine molecules to selfassociate and the counterion to freely interact with the protein surface, with iodide and
thiocyanate forming a very strong interaction, leading to denaturation. These results not only
help to elucidate the arginine mechanism, they also have large implications for interpreting
the mechanism behind the Hofmeister Series (i.e. ion-ion interactions and other ion specific
behaviors are important considerations and the mechanism cannot be completely
generalized). In addition, the results will help to improve the production and formulation of
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.
Schneider et al.
Page 5
protein therapeutics and aid in the development of novel excipients due to a stabilization
method not yet considered.
NIH-PA Author Manuscript
Experimental Methods
Proteins and Reagents
Bovine α-chymotrypsinogen A type II (C4879) and all reagents were obtained from SigmaAldrich (St. Louis, MO) in the highest available grade. All arginine salt forms other than
arginine hydrochloride (ArgHCl) were prepared by titrating an L-arginine solution or
exchanging the chloride ion of ArgHCl. The concentration of aCgn was determined
spectrophotometrically with a PerkinElmer Lambda 35 UV/Vis spectrometer using an
extinction coefficient of 1.97 mL*mg−1cm−1 at 282 nm.56
Accelerated Protein Aggregation
NIH-PA Author Manuscript
The aggregation of α-chymotrypsinogen A (aCgn) was accelerated by incubating samples at
an elevated temperature. aCgn was selected as a model protein because it is a well
characterized protein that has been used extensively as a model protein in aggregation
studies due to it having aggregation characteristics similar to therapeutically relevant
proteins.57 A temperature of 52.5°C was found to be optimal, in that it was well below the
onset of unfolding but accelerated aggregation enough to allow an aggregation experiment
to be completed in less than a day. In addition to all aggregation experiments being
conducted at that temperature, all samples contained 10 mg/mL aCgn, held at pH 5.0 using a
20 mM sodium citrate buffer. Monomer loss was monitored via size exclusion
chromatography using an Agilent 1200 series HPLC, equipped with a Zorbax GF-250
(4.6×250 mm, 4 micron) size exclusion column and a UV-Vis detector. The HPLC mobile
phase contained 25 mM citric acid monohydrate, 25 mM sodium acetate, and 200 mM
sodium chloride with the pH adjusted to 4.0 using sodium hydroxide. The flowrate was set
at 1 mL/min and samples were monitored at a wavelength of 280 nm. For all aCgn
concentrations studied, the monomer loss rate exhibited 2nd order kinetics, thus rate
constants were determined by fitting each data set to a 2nd order rate law.
Preferential Interaction Coefficient Values from VPO Measurements
Preferential interaction theory is the thermodynamic framework to quantify the effect of
cosolutes on protein stability. The preferential interaction coefficient, expressed as either Γμ3
or Γ23 depending on use, is a measure of the preference cosolutes have for the protein
surface and is defined by the following expression,
NIH-PA Author Manuscript
(1)
where m, T, P and μ represent molal concentration, temperature, pressure, and chemical
potential, respectively.58 The subscripts used indicate solution components in Scatchard
notation: water (subscript 1), the protein (subscript 2), and the cosolute (subscript 3).59 We
have previously used vapor pressure osmometry (VPO) to determine the preferential
interaction coefficient of ArgHCl.15 In that article, we explain the implications of this
parameter and how to obtain values utilizing the VPO methodology, thus complete details of
the procedure will not be repeated here. In this report, a Wescor 5520 vapor pressure
osmometer was utilized to determine the osmolality of three separate series (16 samples in
each series) of carefully prepared solutions (i.e. cosolute only solutions, protein only
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.
Schneider et al.
Page 6
solutions, and protein-cosolute solutions at a constant protein concentration), the goal being
to determine the slope with respect to concentration.
NIH-PA Author Manuscript
Differential Scanning Calorimetry
The thermodynamic stability of aCgn in the presence of the arginine salts was determined by
DSC (Microcal VP-Differential Scanning Calorimeter). aCgn was analyzed at a
concentration of 1 mg/mL in a 20 mM sodium citrate pH 5 buffer containing the cosolute of
interest and a scan rate of 90°C/hour. For each arginine salt, three concentrations (including
zero concentration) were analyzed and denaturation midpoint temperature values, Tm, with
respect to cosolute concentration were fitted to a linear trend.
Molecular Dynamics Simulation Methodology
NIH-PA Author Manuscript
Molecular dynamic simulations of protein-cosolute solutions and cosolute only solutions
were performed to give a molecular perspective on how arginine interacts with aCgn, how
arginine interacts with itself, and how the counterion influences these interactions. All
simulations were performed using the NAMD package, with the CHARMM22 force
field.60,61 The TIP3P water model was used.62 The force field parameters for arginine were
taken from the CHARMM force field with the N terminal and the side chain protonated, and
C terminal deprotonated. The parameters for the N and C termini were taken from the CTER
and NTER parameters available in CHARMM. Force field parameters for sulfate, citrate and
thiocyanate ions were taken from the literature.63–65 Force field parameters for acetate,
chloride and dihydrogen phosphate ions were available in CHARMM. Fluoride, bromide
and iodide were not computationally studied due to the absence of reliable force field
parameters in CHARMM. For the purposes of understanding the ordering of anions in the
Hofmeister series, the two opposite cases of sulfate and thiocyanate, with few anions in
between the two extremes, were studied.
Osmotic Virial Coefficient Values from MD Simulations
To characterize the nature of interaction between specific ion pairs, the osmotic second
virial coefficient was calculated from MD simulations by integrating the Mayer function:
(2)
NIH-PA Author Manuscript
The McMillan-Mayer B22 parameter is obviously not the same as the Pitzer Bφ parameter
determined from VPO measurements. However, both parameters describe binary
intermolecular interactions that lead to nonideal osmotic coefficient behavior, thus it is
expected that the trends in both parameters will be identical. The finite size of the simulation
box leads to erroneous normalization of a simulated radial distribution function (RDF).66
Therefore, the asymptotic behavior of the simulated RDF’s are corrected by introducing a
normalization factor, fij (ρ),
(3)
The procedure used for the normalization of RDF’s is included in the supporting
information. The same technique has been used for estimating the osmotic second virial
coefficient of monoatomic salts such as NaCl, KF, etc. Once the correct normalization factor
is obtained, eq 2 is used to estimate the osmotic second virial coefficient for individual ion
pairs using the normalized RDF. The upper limit of the integral was set to a finite cut-off
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.
Schneider et al.
Page 7
NIH-PA Author Manuscript
distance (dependent on the system). The cut-off value was chosen as the distance around
which the osmotic second virial coefficient becomes constant or the corresponding RDF
reach a value close to 1. The cut-off values lie within the range of 1.2–2 nm, with the sulfate
salt having the highest cut-off value.
Preferential Interaction Coefficient Values from MD Simulations
The method for calculating preferential interaction parameters based on a statistical
mechanical method applied to an all-atom model with no adjustable parameter was
developed by Baynes and Trout.67 The number of cosolute and water molecules as a
function of distance from the protein surface is used to calculate Γ23(r,t), where Γ23 is an
alternate and common way to represent Γμ3, until it approaches a constant value. The MD
simulation is saved at periodic time intervals and these saved frames are used to find Γ23(r).
The expression used to calculate the coefficient is 21,68,69
(4)
NIH-PA Author Manuscript
where n3 is the total number of cosolute molecules, and n1 is the total number of water
molecules. The extent of the local domain around a protein in which Γ23(r,t) varies is found
to be around 6 Å in our earlier work. 21,22,65,67,68 The predicted value of the preferential
interaction coefficient Γ23(r = 6 Å) is the average of the instantaneous Γ23(r,t) values. For
arginine salts, which are electrolytic, preferential interaction coefficient calculations would
require an estimate of Γ23 for both the cation and the anion. The preferential interaction
coefficient for a binary electrolyte is given by
(5)
where Γ2,−3 is the preferential interaction coefficient for the anion, Γ2,+3 is the preferential
interaction coefficient for the cation, and ν+/ν− is the ratio of the number of cations and
anions in the chemical formula of the salt. The net charge of the protein divided by the
valency of the ion with the same charge is also subtracted from the preferential interaction
coefficient of the corresponding ion.
NIH-PA Author Manuscript
Results and Discussion
Aggregation Suppression
An example of the data collected during an accelerated aggregation experiment is shown in
Figure 1A. In the figure, the aCgn monomer loss profiles for four different arginine salt
solutions are shown, all at a concentration of 150 mM, in addition to the reference profile
(i.e. no cosolute). It can be seen that ArgH(H2PO4), ArgH(Citrate)1/2, and ArgHCl reduce
the rate of aggregation, while ArgHSCN increases it. The rate of monomer loss in the
presence of ArgHCl at 150 mM is approximately twice as slow as the reference monomer
loss rate. However, at the same concentration, ArgH(Citrate)1/2 and ArgH(H2PO4) slow the
rate by a factor of approximately 3 and 8, respectively. On the other hand, ArgHSCN at 150
mM increases the rate by a factor of 6. This, however, only represents a fraction of the data
collected, which is best represented in Figure 1B. The figure shows the relative rate constant
for monomer loss (i.e. the rate constant relative to the rate constant for no cosolute) over a
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.
Schneider et al.
Page 8
wide range of cosolute concentrations. It should be noted that the profile for ArgH(H2PO4)
is limited due to its poor solubility (a limit of approximately 0.45 M at room temperature).
NIH-PA Author Manuscript
The results shown in Figure 1B demonstrate that dihydrogen phosphate, sulfate, and citrate
significantly improve the aggregation suppression ability of arginine when compared to
chloride. More importantly, at low concentrations (less than 150 mM), the aggregation
suppression ability of ArgH(H2PO4) is significant. At a concentration of 150 mM, the rate of
aCgn aggregation in the presence of ArgH(H2PO4) is 2.5 times slower than when in the
presence of either ArgH(SO4)1/2 or ArgH(Citrate)1/2. However, the aggregation suppression
ability of ArgH(H2PO4) levels off significantly as the concentration is increased beyond 200
mM and the aggregation suppression ability of ArgH(SO4)1/2 and ArgH(Citrate)1/2 likely
exceeds it at concentrations above 300 mM. As will be discussed below, this trend seen for
ArgH(H2PO4) is likely the result of the strong attractive interaction between the
guanidinium moiety of arginine and the phosphate ion. These interactions seem to cause
arginine-phosphate clusters to form at higher concentrations, leading to high preferential
exclusion and a strong thermodynamic stabilization but also limited cosolute solubility and
reduced association suppression.
NIH-PA Author Manuscript
As for the remaining arginine salts to the left of chloride in the Hofmeister Series, the
relative rate constant for fluoride and acetate are essentially the same as that for chloride. At
concentrations below 100 mM, the aggregation suppression ability of these three salt forms
are comparable to sulfate and citrate. However, as the concentration increases, the
aggregation suppression ability of sulfate and citrate continue to improve, having a relative
rate constant value between 0.02–0.04 at the highest concentration tested (0.5 M), while the
relative rate constant for fluoride, acetate, and chloride levels off, having a near constant
value of about 0.2 for concentrations above 0.4 M. ArgHF seems to level off the least and
has a slightly better relative rate constant at a concentration of 0.5 M, which will likely
improve slightly as the concentration is increased further, however, such concentration are
considered quite high and would rarely ever be used.
NIH-PA Author Manuscript
As for the arginine salts to the right of chloride in the Hofmeister Series, the relative rate
constant for bromide remains fairly close to unity for all concentrations tested, while iodide
and thiocyanate have nearly identical profiles, which show a large increase in the rate of
aggregation, up to 10 times faster at a concentration as low as 200 mM. The relative rate
constant for ArgHBr dips down to around 0.7 at 150 mM before it reverses, exceeding a
value of 1 at higher concentrations. This indicates that the slight denaturing effect of
bromide balances out the aggregation suppression effect of arginine, while the strong
denaturing effect of iodide and thiocyanate completely overpower the effect of arginine. The
large size and low charge density of these ions lead to them being poorly hydrated and
therefore, in thermodynamic terms, they tend to partition to the surface of a protein rather
than being solvated in the bulk solution, even more so if hydrophobic residues are
exposed.46 This favorable preferential interaction leads to protein denaturation, with iodide
and thiocyanate having a stronger preferential interaction than bromide due to their smaller
charge density. In fact, iodide and thiocyanate are so destabilizing that ArgHI and ArgHSCN
are stronger denaturants than guanidinium chloride (GdmCl), as represented by relative rate
constant values (see Figure 1) and changes in the denaturation midpoint temperature (see
discussion below). The relative rate constant for GdmCl initially decreases when it is added
to the solution (similar to ArgHBr), reaching a minimum of about 0.6 at 200 mM and then
increases thereafter, reaching a value of 9 at 0.7 M, the highest concentration tested. This
shows that at low concentrations, GdmCl has an ability to reduce aggregation by inhibiting
protein-protein interactions, similar to arginine, but at high concentration, the denaturing
effect dominates and causes aggregation to proceed more rapidly. However, ArgHI and
ArgHSCN show no stabilizing effect from the presence of arginine. They instead increase
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.
Schneider et al.
Page 9
NIH-PA Author Manuscript
the rate of aggregation at all concentrations tested, doubling the rate at a concentration as
low as 50 mM. All of this exemplifies the pronounced effect the choice of anion has on
protein stability and shows that ArgH(H2PO4), ArgH(SO4)1/2 and ArgH(Citrate)1/2 might be
more useful than ArgHCl for preventing protein aggregation.
Conformational Stability
The thermodynamic stability of aCgn in the presence of the arginine salts was assessed by
determining the change in the denaturation midpoint temperature, Tm, with respect to
cosolute concentration. Shifts in the melting temperature of the protein are indicative of
shifts in the folding equilibrium of the protein, with increases in the melting temperature
corresponding to a shift toward the native state, which in return reduces the number of
aggregation prone species. Table I shows the denaturation midpoint temperature increment,
dTm/d[3], defined as the slope of Tm with respect to the molar concentration of the cosolute
(component 3), over the concentration range tested, as determined by DSC. Over this short
initial concentration range, Tm exhibited a linear trend, giving constant values for dTm/d[3].
The values obtained exhibit the pronounced effect the anions have on the thermodynamic
stability of the protein and they correspond directly to the ordering of the relative rate
constant values obtained during the aggregation study.
NIH-PA Author Manuscript
NIH-PA Author Manuscript
As shown in Table I, ArgH(H2PO4), ArgH(SO4)1/2 and ArgH(Citrate)1/2 increase the
melting temperature of the protein with an initial rate of 6.3, 5.4, and 3.3 °C*M−1,
respectively. Clearly, these salts are able to stabilize the native structure of the protein and
inhibit the formation of partially unfolded species. This thermodynamic stabilization is on
top of the association suppression effect of arginine, leading to the significant aggregation
suppression shown in Figure 1. ArgH(Acetate), ArgHF, and ArgHCl exhibit no
thermodynamic stabilization, which is consistent with previous reports for ArgHCl, thus
these salts inhibit aggregation solely through an association suppression mechanism.
ArgHBr has a denaturing effect on aCgn, lowering the melting temperature at a rate of
7.7°C*M−1. This reduction in conformational stability, which increases the number of
reactive species in solutions, seems to nearly balance out the association suppression effect
of arginine, as exhibited by the relative rate constant having a value near unity over a wide
concentration range. This reduction in thermodynamic stability is slightly greater than that
for GdmCl, which might explain why they have near identical profiles at low concentrations.
However, the association suppression effect of arginine is greater than that for guanidinium,
thus, at high concentrations, the rate of aCgn aggregation is higher in the presence of GdmCl
than in the presence of ArgHBr. As for the other arginine salts, ArgHI and ArgHSCN lower
the melting temperature significantly, both with a rate greater than 20 °C*M−1. This
denaturing effect overpowers the association suppression effect of arginine and as a result,
these salts increase the rate of aggregation at all concentrations. These results, along with
aggregation suppression data, support the claim that ArgH+ acts as an association suppressor
and shows that the choice of anion will either enhance or counteract this aggregation
suppression by influencing the stability of the native structure.
Ion-Ion Interactions
As mentioned in the introduction, ion-ion interactions may be contributing to the observed
behavior of the arginine salts. To quantify this behavior, osmotic 2nd virial coefficient values
were determined for aqueous solutions of the arginine salts via VPO and MD simulations.
Experimental osmotic data were analyzed by using the Pitzer model (see supporting
information), while MD simulations were analyzed using the McMillan-Mayer model. Pitzer
ion interaction parameters for each of the salts are shown in Table S2 in the supporting
information, which shows that the second virial coefficient parameters obtained for ArgHCl
closely matches values reported in the literature. Since the magnitude of like and unlike ion
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.
Schneider et al.
Page 10
NIH-PA Author Manuscript
interactions is ionic strength dependent, it is generally held that only cation-anion binary
interactions make contributions to the value of β(1). Both like and unlike ion interactions
make contributions to β(0), making it difficult to interpret this parameter unless radial
distribution functions are available for each ion pair. The trends observed for β(1) are in
accordance with the discussion above, which are highlighted in Table I. The values for the
dihydrogen phosphate, sulfate, and citrate salts are large and negative, indicating a strong
attractive interaction between arginine and those anions. The value for ArgHCl is negative
but smaller in magnitude, indicating a weak attractive interaction. The values for the other
halide salts show repulsive interactions for those salt forms with the strongest repulsion for
thiocyanate and iodide and the weakest for bromide. It should be noted that these trends are
similar to the virial coefficient values for guanidinium halide salts, which show a weak
repulsive interaction with fluoride, a moderate repulsive interaction with bromide, and a
strong repulsive interaction with iodide.70
NIH-PA Author Manuscript
During the simulations, significant ion pairing was observed in the aqueous arginine salt
solutions, which supports the above experimental results. The ion pairing was also observed
to be dependent on the choice of anion. Ion pairs in thiocyanate, chloride and acetate
solutions are observed to be randomly distributed throughout the solution while citrate,
phosphate and sulfate show a marked tendency to form hydrogen bonded clusters. Snapshots
of the MD simulation boxes of arginine salts are shown in Figure 2. It can be seen that the
structures formed by sulfate, citrate and phosphate salts are worm-like chains of sizes
comparable to the box-size. Clustering observed in chloride, thiocyanate and acetate
solutions is mainly due to Arg-Arg interaction as reported in our recent work on aqueous
arginine hydrochloride solutions.22,71 Similar ionic clusters were also observed in aqueous
guanidinium salt solutions.47–49 Brady and coworkers have shown that clustering in Gdm
salts is not dependent on the choice of the water model and is not a simulation artifact. Due
to the presence of Gdm group as a side-chain in arginine, the clustering behavior of arginine
salts is expected to be similar to Gdm salts. The only difference between Gdm and arginine
salts is the presence of additional charged groups and a carbon chain in arginine. These
additional charged groups (N-terminal amino group and C-terminal carboxylate group)
further enhance the clustering in solution.
NIH-PA Author Manuscript
Radial distribution functions (see Figure 3) between ion pairs help to further characterize the
solution structure. Arg-Arg pairing, as illustrated by the RDF between the Gdm carbon
atoms in the arginine molecule (see Figure 3a), shows a tendency of the Gdm side chain to
stack with each other or form ion-pairs separated by the anion acting as a bridge between
two Gdm side chains. The presence of an anion does not significantly affect the Arg-Arg
pairing except in the case of citrate, where the large size of the anion crowds out the
formation of Arg pairs. The pairing between cation and anion is strongest for the sulfate ion
due to the exceptionally strong interaction between the Gdm group and sulfate (see Figure
3b). The acetate ion also has a strong interaction with arginine, however, the strong binding
of sulfate and acetate is in stark contrast to the weak binding of thiocyanate and chloride.
The pairing between anions (see Figure 3c) is significantly weak, as compared to other ion
pairs in solution. Phosphate ion pairing exhibits a peak around 0.5 nm due to the presence of
hydrogen bonding acceptors and donors in the same molecular anion. Sulfate also shows a
peak around 0.6–0.7 nm but this is not due to direct interaction between sulfate ions; rather,
the peak is due to the presence of multiple sulfate anions associating with adjacent or the
same arginine cation. The peak heights for sulfate and chloride salts are comparable to the
peak heights observed by Brady and coworkers for guanidinium salts. They also observe
rugged RDF’s due to the extensive clustering in guanidinium carbonate and sulfate salt
solutions.47–49 The RDF’s between ions give an idea about their clustering behavior but
these plots cannot be used as a direct measure of observed clustering in different salt
solutions. Peak heights and positions are dependent on the charge of the ion, the number and
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.
Schneider et al.
Page 11
NIH-PA Author Manuscript
type of hydrogen-bonding groups, the choice of reference atom for calculation of RDF and
bulk densities. Therefore, second osmotic virial coefficient values provide a better measure
to compare different salts.
NIH-PA Author Manuscript
To elaborate, the individual contribution of specific ion-pairs to the overall clustering in the
solution can be estimated by calculating the osmotic second virial coefficient between ionpairs in solution. B22 values for ion-pairs in each salt solution are reported in Table II. It can
be seen that the contribution of cation-cation interaction is similar in magnitude to cationanion interaction. Σij denotes the sum of interactions between all ion pairs. Sulfate,
dihydrogen phosphate and citrate form a group with strong overall interactions whereas
chloride, thiocyanate and acetate form a group with weak overall interactions. The
difference between the collective structures of these two groups of anions apparently results
from the differences in the hydrogen bonding of the anions to the arginine cation. The
numbers of hydrogen bonds between different ion-pairs in solution are reported in Table III.
It can be seen that the highest number of hydrogen bonds between arginine molecules are
formed in chloride and thiocyanate solutions due to the minimal interaction of these ions
with the cation. The loss of hydrogen bonds between arginine molecules in acetate, sulfate,
phosphate and citrate solutions is compensated by the formation of large number of cationanion hydrogen bonds. Acetate also interacts strongly with the arginine, forming hydrogen
bonds with the Gdm group as shown in Figure 4a, but due to the limited number of
hydrogen bond acceptors, the acetate anion cannot act as a bridge between multiple cations.
On the other hand, citrate, sulfate and dihydrogen phosphate can form bridged structures, as
shown in Figure 4b, 4c, and 4d, respectively. The number of hydrogen bonds between
arginine ions (with an anion acting as a bridge) is high for sulfate, dihydrogen phosphate and
citrate as compared to acetate, chloride and thiocyanate, which have limited to negligible
capacity to form such bridged structures. These observations also explain why sulfate,
dihydrogen phosphate and citrate form large hydrogen bonded clusters in solution as
compared to acetate, chloride and thiocyanate. These results are in accordance to our
previous reports, in which we showed that ArgHCl likely forms Arg-Arg clusters in solution
at high concentration, which leads to a higher preferential exclusion.
Preferential Interactions
NIH-PA Author Manuscript
To gain insight into how the arginine salts inhibit protein-protein interactions, the
preferential interaction coefficient, Γμ3 or Γ23, at various concentrations was determined,
both experimentally via VPO measurements and computationally via MD simulations. The
experimental results are depicted in Figure 5, which shows Γμ3 values at various cosolute
concentrations, and Table S2, which summarizes the polynomial fit and uncertainty of the
experimental data. It should be noted that the preferential interaction coefficient could not be
obtained via VPO for the acetate and fluoride salts due to these anions being in equilibrium
with their respective conjugate acid forms at pH 5. This is because both of these acid forms
are volatile and this volatility (though slight due to the low concentration of the conjugate
acid) interfered with the VPO measurements. Regardless, the experimental results for the
other salt show a clear difference between the anions to the left and to the right of chloride
in the Hofmeister Series. The chloride, bromide, iodide, and thiocyanate salts all have
preferential interaction coefficient values that start out positive at low concentrations, with
an initial slope almost identical to GdmCl, whereas the initial values for the citrate, sulfate,
and dihydrogen phosphate salts are negative. The preferential interaction data for GdmCl is
depicted in Figure 5 and Table I for comparison, which exhibits a linear trend with respect to
concentration.
The initial slope for the chloride, bromide, iodide, and thiocyanate salts show that at low
concentrations, when there is likely little ion-ion interactions, the salts interact favorably
with the protein surface, with an attractive interaction similar in strength as that for GdmCl.
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.
Schneider et al.
Page 12
NIH-PA Author Manuscript
However, as the concentration increases, the preferential interaction coefficient values for all
of these arginine salts begin to decrease and ultimately become negative, whereas the trend
for GdmCl continues to increase linearly, indicating that at high concentrations, these
arginine salts exhibit a net repulsive interaction with the native protein structure, opposite of
what is observed for GdmCl. Furthermore, the preferential interaction coefficient values for
thiocyanate and iodide salts exhibited the largest values at all concentrations, followed by
bromide and chloride, indicating that the thiocyanate and iodide ions cause those salts to
have a stronger attractive interaction with the protein surface, followed by bromide, which
causes only a slightly stronger interaction.
NIH-PA Author Manuscript
This nonlinear trend is in accordance with previously reported values for ArgHCl and aCgn,
with values differing slightly due to a difference in pH.15 In our previous report, the
surprising behavior exhibited by ArgHCl was initially thought to be the result of the protein
surface quickly becoming saturated with the large molecule. MD simulations later revealed
that this hypothesis was likely true but only at much higher concentrations (>1.5 m) and thus
another mechanism contributed to the nonlinear trend observed at lower concentrations.22,32
As mentioned before, studies of aqueous ArgHCl solutions revealed that arginine has a
strong propensity to form clusters in solution via hydrogen bonding between the
guanidinium and carboxylate moieties, which was found to be stronger than the hydrogen
bond formed with water, and as a result, it reduces the interaction with the protein surface.
This reduced interaction is due, in part, to the larger net size of the clusters (steric exclusion)
but in large part, is due to a favorable interaction with other arginine molecules in the bulk
solution, which is similar or greater in strength as the interaction with the protein surface.
This self-interaction is concentration dependent, with more clustering obviously occurring at
higher concentrations, leading to the nonlinear trend shown in Figure 5.
NIH-PA Author Manuscript
It should be noted that preferential interaction coefficient measurements and the interactions
studied via MD simulations are for interactions with the native protein structure only and are
not necessarily indicative of the propensity to denature protein molecules. However, most
cosolutes that are attracted to the native state are typically more attracted to the unfolded
state and vice versa for preferentially excluded cosolutes, due to most interactions being
nonspecific in nature. It is this difference in the preferential interaction between the two
states that leads to a change in the thermodynamic stability of the protein, which can be
represented by the Wyman linkage relationship. Therefore, even though for most cosolutes
there is a direct relationship between the preferential interaction coefficient and the folding
stability of the protein, there are, however, some exceptions.72 The mechanism for each
exception varies but all are the result of preferential interactions being repulsive in one state
and attractive in the other. For compounds containing guanidinium and halide anions,
exceptions can arise due to the exposure of hydrophobic residues upon unfolding.
Guanidinium, the larger halides, and thiocyanate are poorly hydrated ions that have a strong
preference for hydrophobic residues which causes them to interact more strongly with the
unfolded state and as a result, they denature proteins.46 That is, if they are not inhibited from
interacting with the protein surface, as in the case for arginine. The preferential interaction
coefficient values obtained experimentally and MD simulations seem to indicate that
arginine is inhibited from interacting strongly with the protein (either with the native state or
the unfolded state) even though it has a guanidinium moiety, but the halide anions and
thiocyanate are free to interact with the protein. This favorable interaction is in order of
SCN− ≈ I− > Br− > Cl−, as indicated by Tm and Γμ3 results. The interaction between F− and
the protein is likely similar in strength as that for Cl−, which is weak at best, but as
explained above, Γμ3 could not be determined for ArgHF.
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.
Schneider et al.
Page 13
NIH-PA Author Manuscript
The magnitude of the interaction with the protein for these salts not only arises from a
difference in the protein-anion interaction but also from a difference in the guanidiniumanion interaction, as discussed above, with thiocyanate and iodide exhibiting a weaker
interaction with guanidinium. Thus these ions are less likely to interact with arginine in the
bulk solution. Fluoride and chloride not only exhibit a weaker interaction with the protein
surface, but they also interact more favorably with the guanidinium functional group on
arginine. Therefore, as a result of Arg-Arg interactions and weak Arg-anion interactions, Γμ3
values exhibit a nonlinear trend for all of the salts discussed and due to the variation in Arganion and protein-anion interactions, the magnitude of Γμ3 and the shift in Tm varies
amongst the salts. ArgHCl and ArgHF are neither naturants nor denaturants because there is
little difference between the interaction with the native and unfolded states. When a protein
unfolds, free ArgH+ will likely interact more strongly with the exposed hydrophobic
residues but the clusters should become more excluded, counteracting this interaction. The
interaction of the fluoride and chloride ions with the unfolded state likely does not change
much because of the weak interaction these ions have for hydrophobic residues. As for
ArgHBr, ArgHI, and ArgHSCN, the interaction with the unfolded state is stronger due to the
stronger attractive interaction the anions exhibit for hydrophobic residues, leading to those
salts denaturing proteins.
NIH-PA Author Manuscript
The salts to the left of chloride in the Hofmeister series exhibit similar interactions but of
differing magnitude, resulting in Γμ3 trends distinct from the other salts. The dihydrogen
phosphate, sulfate, and citrate salts have Γμ3 values that start out negative and remain
negative, with values similar to other highly excluded solutes (e.g. glycerol), indicating that
the salts have a strong repulsive interaction with the protein at all concentrations.15
ArgH(SO4)1/2 is the least excluded of these salts, causing the ordering of Γμ3 data to differ
slightly from the ordering of Tm data. Also, the profile for citrate is almost linear for most of
the concentration range, likely from the large citrate ion being highly excluded and being a
hindrance to Arg-Arg interactions. It is well documented that salts with these anions
typically stabilize the folded state, as is the case for the data shown here. Both this and the
preferential exclusion can be contributed, in part, to the nonspecific repulsive interaction of
the anions with protein molecules that result from the ions being well hydrated and
preferring the bulk solution. But as discussed above, dihydrodgen phosphate, sulfate and
citrate have been shown to exhibit a strong attractive interaction with guanidinium, which
further reduces the interaction with the protein.
NIH-PA Author Manuscript
Theoretical preferential interaction coefficient values computed from the MD simulation are
reported in Table IV. The estimated preferential interaction values match well with the trend
in the experimental values. The experimental and computational estimates for sulfate,
dihydrogen phosphate (if linearly extrapolated to 0.5m) and citrate indicate large, negative
preferential interaction coefficient values, while thiocyanate exhibits a large, positive
preferential interaction coefficient value. The experimental and computational estimates do
not match exactly, which is likely due to the inability of the force field to capture the
interactions of these ions with the protein surface. Force fields are optimized for interaction
of ions with water molecules. However, as shown above, for arginine salts, interactions with
water is only one of the key interactions. In order to obtain an accurate estimate of
preferential interaction coefficient values, force fields for electrolytes should also be
benchmarked to reproduce bulk solution properties and interactions between individual
ions.73,74 For the purposes of understanding the interesting experimental trends observed in
this study, the current force fields provide reasonable estimates of the trends in the data. On
the basis of the preferential interaction coefficient of individual ions, it can be concluded
that the number of cations on the protein surface is related to the number of anions.
Thiocyanate ions have high positive preferential interaction coefficient values but it does not
interact strongly with arginine cations. Sulfate ions have a negative preferential interaction
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.
Schneider et al.
Page 14
NIH-PA Author Manuscript
coefficient but it interacts strongly with arginine. Therefore, there are two driving forces that
push and pull arginine towards and away from the protein surface. In order to conserve the
local charge near the protein surface, more arginine molecules are drawn towards the protein
surface if there are a high number of anions near the protein surface. If the interactions
between ions in the bulk solution are not as strong as compared to the interaction between
the ion and the protein surface, the ions accumulate near the protein surface. Therefore,
sulfate, phosphate and citrate salts have a large negative Γμ3 because of their exclusion from
the protein surface and the strong interaction with arginine molecules and vice-versa for
thiocyanate, acetate and chloride salts.
This interaction between arginine and the anion further reduces the preferential interaction
of the arginine salt with the surface of the protein and further limits interactions with the
unfolded state, thus enhancing thermodynamic stabilization. This cation-anion interaction
does not necessarily reduce Arg-Arg interactions though, because the anion bridges together
two arginine molecules, similar to guanidinium-sulfate clusters discovered by Mason and
coworkers, causing the nonlinear trend observed.48 Therefore, the ordering of the Γμ3 and
Tm data is not only ranked by how the individual anions interact with protein molecules but
also, with how those anions interact with the guanidinium functional group on the arginine
molecule.
NIH-PA Author Manuscript
Mechanistic Insight
NIH-PA Author Manuscript
The trends discussed here are related to the impact these salts have on protein aggregation.
As discussed in the previous section, both changes in the folding equilibrium of the protein
and changes in protein-protein interactions must be taken into consideration. The halide and
thiocyanate salts exhibit an attractive interaction with the protein at low concentrations.
This, according to the theory developed by Baynes and Trout, should reduce protein-protein
interactions, much more so for arginine than for guanidinium due to the larger size of
arginine.14,75 However, as discussed before, bromide, iodide, and thiocyanate also denature
the protein, which counteracts this effect. This is not the case for chloride and fluoride, thus
ArgHCl and ArgHF reduce aggregation solely by reducing protein-protein interactions.
However, at high concentrations, the ArgHCl shows a net exclusion (which is likely the case
for ArgHF as well but to a lesser extent) and the aggregation suppression ability of ArgHCl
plateaus because adding additional ArgHCl molecules at that point does not contribute any
to protein-protein interaction suppression due to the additional arginine molecules being
partitioned into the bulk solution. Even though ArgHCl shows a net exclusion at high
concentrations, there are enough arginine molecules in the local domain of the protein that it
significantly reduces protein-protein interactions, thus lowering the relative rate constant to
about 0.2. This indicates that the effect from the arginine molecule should be significant for
the highly excluded salt forms as well, though the effect of arginine might be reduced some.
Furthermore, the aggregation suppression is further enhanced by the stabilizing effect the
citrate, sulfate, and dihydrogen phosphate salt forms have on the folding equilibrium,
leading to the relative rate constant being further reduced beyond that for ArgHCl.
Moreover, another contributing factor may be an increased diffusional barrier. The cluster
formation induced by the phosphate, sulfate, and citrate ions likely increases the viscosity of
the solution in addition to forming a network of bridged ions around the protein, thus
inhibiting the movement of protein molecules.
The clustering due to the presence of hydrogen-bonding anions leads to the formation of
large clusters, which are either present within the local domain surrounding the protein or
attached to the protein surface via hydrogen-bonds. These clusters of hydrogen-bonded
molecules would be difficult to remove from the surface as compared to single molecules
within the local domain. In earlier studies, we shown that arginine self-interaction limits the
binding of guanidinium group to the protein surface, which explained why arginine is not a
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.
Schneider et al.
Page 15
NIH-PA Author Manuscript
protein destabilizer despite the presence of a denaturing guanidinium group.32 In this study,
we have shown the effect of enhanced self-interaction in aqueous arginine salt solutions (due
to the use of hydrogen-bonding anions) on protein-protein association.
Conclusions
NIH-PA Author Manuscript
For arginine salt solutions, we have shown that the interaction between ions and the
interaction of the anion with the protein surface influences the preferential interaction
coefficient value, which is directly related to the conformational and colloidal stability of the
protein. Attractive ion-ion interactions in solution lead to the formation of clusters that are
larger in size as compared to the individual ions. These large clusters should be more
effective at decreasing protein-protein association than small molecules. Furthermore,
clustering should lead to an increase in the viscosity of the solution, which would lower the
diffusion of proteins in solution. Therefore, ion-ion interactions affect the rate of
aggregation via three mechanisms: 1) enhancing the native state conformational stability, 2)
increasing the barrier for protein-protein association, and 3) decreasing protein-protein
encounters by increasing the viscosity of the solution medium. These results not only help to
elucidate the arginine mechanism, they also have large implications for interpreting the
mechanism behind the Hofmeister Series, in that changes in the water structure did not seem
to come into play. Rather, ion-ion interactions and other ion specific behavior seem to
dominate the behavior of each arginine salt. On a last note, for a more definitive picture of
how arginine influences aggregation, the research conducted here should be extended to
other proteins and possibly a wider range of pH to better understand what other biophysical
effects contribute to the arginine mechanism.
Supplementary Material
Refer to Web version on PubMed Central for supplementary material.
Acknowledgments
The authors acknowledge funding support from the National Institutes of Health and the Singapore-MIT Alliance.
References
NIH-PA Author Manuscript
1. Wang W. Int J Pharm. 2005; 289:1–30. [PubMed: 15652195]
2. Manning MC, Chou DK, Murphy BM, Payne RW, Katayama DS. Pharm Res. 2010; 27:544–575.
[PubMed: 20143256]
3. Cromwell MEM, Hilario E, Jacobson F. AAPS J. 2006; 8:E572–E579. [PubMed: 17025275]
4. Rosenberg AS. AAPS J. 2006; 8:E501–E507. [PubMed: 17025268]
5. Randolph TW, Carpenter JF. AIChE J. 2007; 53:1902–1907.
6. Leader B, Baca QJ, Golan DE. Nat Rev Drug Discovery. 2008; 7:21–39.
7. Frokjaer S, Otzen DE. Nat Rev Drug Discovery. 2005; 4:298–306.
8. Cleland JL, Powell MF, Shire SJ. Crit Rev Ther Drug Carrier Syst. 1993; 10:307–377. [PubMed:
8124728]
9. Wang W. Int J Pharm. 1999; 185:129–188. [PubMed: 10460913]
10. Wang W, Singh S, Zeng DL, King K, Nema S. J Pharm Sci. 2007; 96:1–26. [PubMed: 16998873]
11. Jorgensen L, Hostrup S, Moeller EH, Grohganz H. Expert Opin Drug Deliv. 2009; 6:1219–1230.
[PubMed: 19678792]
12. Lange C, Rudolph R. Curr Pharm Biotechnol. 2009; 10:408–414. [PubMed: 19519416]
13. Buchner J, Rudolph R. Bio-Technology. 1991; 9:157–162. [PubMed: 1369317]
14. Baynes BM, Wang DIC, Trout BL. Biochemistry. 2005; 44:4919–4925. [PubMed: 15779919]
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.
Schneider et al.
Page 16
NIH-PA Author Manuscript
NIH-PA Author Manuscript
NIH-PA Author Manuscript
15. Schneider CP, Trout BL. J Phys Chem B. 2009; 113:2050–2058. [PubMed: 19199688]
16. Arakawa T, Tsurnoto K, Nagase K, Ejima D. Protein Expr Purif. 2007; 54:110–116. [PubMed:
17408966]
17. Das U, et al. PLoS One. 2007; 2:e1176. [PubMed: 18000547]
18. Srinivas V, Raman B, Rao KS, Ramakrishna T, Rao CM. Protein Sci. 2003; 12:1262–1270.
[PubMed: 12761397]
19. Ghosh R, Sharma S, Chattopadhyay K. Biochemistry. 2009; 48:1135–1143. [PubMed: 19146403]
20. KRCR, Lilie H, Rudolph R, Lange C. Protein Sci. 2005; 14:929–935. [PubMed: 15741330]
21. Shukla D, Shinde C, Trout BL. J Phys Chem B. 2009; 113:12546–12554. [PubMed: 19697945]
22. Shukla D, Trout BL. J Phys Chem B. 2010; 114:13426–13438. [PubMed: 20925358]
23. Arakawa T, Ejima D, Tsumoto K, Obeyama N, Tanaka Y, Kita Y, Timasheff SN. Biophys Chem.
2007; 127:1–8. [PubMed: 17257734]
24. Nakakido M, Kudou M, Arakawa T, Tsumoto K. Curr Pharm Biotechnol. 2009; 10:415–420.
[PubMed: 19519417]
25. Arakawa T, Kita Y, Koyama AH. Int J Pharm. 2008; 355:220–223. [PubMed: 18242019]
26. Hirano A, Tokunaga H, Tokunaga M, Arakawa T, Shiraki K. Arch Biochem Biophys. 2010;
497:90–96. [PubMed: 20346908]
27. Ohtake S, Arakawa T, Koyama AH. Molecules. 2010; 15:1408–1424. [PubMed: 20335989]
28. Arakawa T, Hirano A, Shiraki K, Kita Y, Koyama AH. Int J Biol Macromol. 2010; 46:217–222.
[PubMed: 19948185]
29. Abe R, Kudou M, Tanaka Y, Arakawa T, Tsumoto K. Biochem Biophys Res Commun. 2009;
381:306–310. [PubMed: 19167355]
30. Tsumoto K, Umetsu M, Kumagai I, Ejima D, Philo JS, Arakawa T. Biotechnol Prog. 2004;
20:1301–1308. [PubMed: 15458311]
31. Arakawa T, Tsumoto K. Biochem Biophys Res Commun. 2003; 304:148–152. [PubMed:
12705899]
32. Shukla D, Trout BL. J Phys Chem B. 2011; 115:1243–1253. [PubMed: 21186800]
33. Hofmeister F. Arch Exp Pathol Pharmakol. 1888; 24:247–260.
34. Baldwin RL. Biophys J. 1996; 71:2056–2063. [PubMed: 8889180]
35. Cacace MG, Landau EM, Ramsden JJ. Q Rev Biophys. 1997; 30:241–277. [PubMed: 9394422]
36. Curtis RA, Ulrich J, Montaser A, Prausnitz JM, Blanch HW. Biotechnol Bioeng. 2002; 79:367–
380. [PubMed: 12115400]
37. Zhang YJ, Cremer PS. Curr Opin Chem Biol. 2006; 10:658–663. [PubMed: 17035073]
38. Tadeo X, Pons M, Millet O. Biochemistry. 2007; 46:917–923. [PubMed: 17223714]
39. Lodderstedt G, Sachs R, Faust J, Bordusa F, Kuhn U, Golbik R, Kerth A, Wahle E, Balbach J,
Schwarz E. Biochemistry. 2008; 47:2181–2189. [PubMed: 18205394]
40. Maity H, Karkaria C, Davagnino J. Curr Pharm Biotechnol. 2009; 10:609–625. [PubMed:
19619121]
41. Golovanov AP, Hautbergue GM, Wilson SA, Lian L-Y. J Am Chem Soc. 2004; 126:8933–8939.
[PubMed: 15264823]
42. Valente JJ, Verma KS, Manning MC, Wilson WW, Henry CS. Biophys J. 2005; 89:4211–4218.
[PubMed: 16199499]
43. Ishibashi M, Tsumoto K, Tokunaga M, Ejima D, Kita Y, Arakawa T. Protein Expr Purif. 2005;
42:1–6. [PubMed: 15893471]
44. Samuel D, Kumar TKS, Ganesh G, Jayaraman G, Yang PW, Chang MM, Trivedi VD, Wang SL,
Hwang KC, Chang DK, Yu C. Protein Sci. 2000; 9:344–352. [PubMed: 10716186]
45. Ignatova Z, Gierasch LM. Proc Natl Acad Sci U S A. 2006; 103:13357–13361. [PubMed:
16899544]
46. Mason PE, Neilson GW, Dempsey CE, Barnes AC, Cruickshank JM. Proc Natl Acad Sci U S A.
2003; 100:4557–4561. [PubMed: 12684536]
47. Mason PE, Neilson GW, Enderby JE, Saboungi ML, Dempsey CE, MacKerell AD, Brady JW. J
Am Chem Soc. 2004; 126:11462–11470. [PubMed: 15366892]
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.
Schneider et al.
Page 17
NIH-PA Author Manuscript
NIH-PA Author Manuscript
NIH-PA Author Manuscript
48. Mason PE, Dempsey CE, Neilson GW, Brady JW. J Phys Chem B. 2005; 109:24185–24196.
[PubMed: 16375411]
49. Mason PE, Neilson GW, Kline SR, Dempsey CE, Brady JW. J Phys Chem B. 2006; 110:13477–
13483. [PubMed: 16821873]
50. Dempsey CE, Mason PE, Brady JW, Neilson GW. J Am Chem Soc. 2007; 129:15895–15902.
[PubMed: 18052164]
51. Mason PE, Dempsey CE, Vrbka L, Heyda J, Brady JW, Jungwirth P. J Phys Chem B. 2009;
113:3227–3234. [PubMed: 19708168]
52. Bonner OD. J Chem Thermodyn. 1976; 8:1167–1172.
53. Bonner OD. J Phys Chem. 1977; 81:2247–2249.
54. Schlund S, Schmuck C, Engels B. Chem-Eur J. 2007; 13:6644–6653.
55. Woods AS, Ferre S. J Proteome Res. 2005; 4:1397–1402. [PubMed: 16083292]
56. Gekko K, Timasheff SN. Biochemistry. 1981; 20:4667–4676. [PubMed: 7295639]
57. Andrews JM, Roberts CJ. Biochemistry. 2007; 46:7558–7571. [PubMed: 17530865]
58. Courtenay ES, Capp MW, Anderson CF Jr, MTR. Biochemistry. 2000; 39:4455–4471. [PubMed:
10757995]
59. Scatchard G. J Am Chem Soc. 1946; 68:2315–2319. [PubMed: 21002231]
60. Phillips JC, Braun R, Wang W, Gumbart J, Tajkhorshid E, Villa E, Chipot C, Skeel RD, Kale L,
Schulten K. J Comput Chem. 2005; 26:1781–1802. [PubMed: 16222654]
61. Brooks BR, Bruccoleri RE, Olafson BD, States DJ, Swaminathan S, Karplus M. J Comput Chem.
1983; 4:187–217.
62. Jorgensen WL, Chandrasekhar J, Madura JD, Impey RW, Klein ML. J Chem Phys. 1983; 79:926–
935.
63. Cannon WR, Pettitt BM, McCammon JA. J Phys Chem. 1994; 98:6225–6230.
64. Sansone R, Ebner C, Probst M. J Mol Liq. 2000; 88:129–150.
65. Shukla D, Zamolo L, Cavallotti C, Trout BL. J Phys Chem B. 2011; 115:2645–2654. [PubMed:
21355601]
66. Lyubartsev AP, Marcelja S. Phys Rev E. 2002; 65
67. Baynes BM, Trout BL. J Phys Chem B. 2003; 107:14058–14067.
68. Vagenende V, Yap MGS, Trout BL. J Phys Chem B. 2009; 113:11743–11753. [PubMed:
19653677]
69. Kang M, Smith PE. Fluid Phase Equilib. 2007; 256:14–19.
70. Das B. J Solut Chem. 2004; 33:33–45.
71. Vondrasek J, Mason PE, Heyda J, Collins KD, Jungwirth P. J Phys Chem B. 2009; 113:9041–
9045. [PubMed: 19354258]
72. Timasheff SN. Annu Rev Biophys Biomolec Struct. 1993; 22:67–97.
73. Kalcher I, Dzubiella J. J Chem Phys. 2009; 130
74. Fyta M, Kalcher I, Dzubiella J, Vrbka L, Netz RR. J Chem Phys. 2010; 132
75. Baynes BM, Trout BL. Biophys J. 2004; 87:1631–1639. [PubMed: 15345542]
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.
Schneider et al.
Page 18
NIH-PA Author Manuscript
Figure 1.
NIH-PA Author Manuscript
The influence of arginine salts on aCgn monomer loss due to aggregation at 52.5°C. All
solutions contained 10 mg/mL aCgn and were prepared in a 20 mM sodium citrate pH 5
buffer. (A) Monomer concentration, M, normalized with respect to the initial monomer
concentration, M0, versus time from a single experiment, with plots fitted to a second order
rate law. The concentration of each arginine salt shown was 150 mM. (B) aCgn monomer
loss rate constant with a cosolute present, k, relative to no cosolute, k0, versus cosolute
concentration, with lines drawn through the plots to aid the eye.
NIH-PA Author Manuscript
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.
Schneider et al.
Page 19
NIH-PA Author Manuscript
NIH-PA Author Manuscript
NIH-PA Author Manuscript
Figure 2.
Snapshots of the MD simulation box containing arginine salts at a molal concentration of 0.5
mol/kg. To improve the clarity of the image, water molecules are not shown and only heavy
atoms (all atoms excluding hydrogen) in the arginine molecules and counterions are shown.
The following color code is used to represent atoms: C (cyan), O (red), N (blue), S (yellow),
Cl (light blue), and P (brown). Arginine molecules are shown in silver.
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.
Schneider et al.
Page 20
NIH-PA Author Manuscript
NIH-PA Author Manuscript
NIH-PA Author Manuscript
Figure 3.
Radial distribution functions (RDF) between ion pairs in aqueous arginine salt solutions.
The Cation-Cation RDF is the RDF between the guanidinium carbon atoms of arginine. For
the counterions, the atoms used as centers for estimating the RDF’s are: Sulfate – Sulfur
atom, Phosphate – Phosphorus atom, Citrate – Central carbon atom, Thiocyanate – Nitrogen
atom and Acetate – Carboxylate carbon atom.
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.
Schneider et al.
Page 21
NIH-PA Author Manuscript
NIH-PA Author Manuscript
Figure 4.
Hydrogen bonding interaction between arginine and (a) acetate, (b) citrate, (c) sulfate and
(d) phosphate anions. It can be seen that sulfate, phosphate and citrate can interact with
multiple arginine molecules forming large hydrogen-bonded structures. The following color
code is used to represent atoms: C (cyan), O (red), N (blue), S (yellow), and P (brown).
NIH-PA Author Manuscript
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.
Schneider et al.
Page 22
NIH-PA Author Manuscript
NIH-PA Author Manuscript
Figure 5.
Preferential Interaction Coefficient, Γμ3, values for the interaction between arginine salts
(and guanidinium chloride for comparison) and aCgn, as determined from VPO
measurements. Error bars left off for clarity and curves drawn through the plots to aid the
eye (see Table S2 for more detail).
NIH-PA Author Manuscript
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.
Schneider et al.
Page 23
Table I
NIH-PA Author Manuscript
Summary of aCgn denaturation midpoint temperature increment and Pitzer ion interaction parameter for
arginine salts and guanidinium chloride.
Salt
dTm/d[3] K*L/mol
Max. [3] mol/L
β(1) kg/mol
ArgH(H2PO4)
6.3
0.2
−1.5303
ArgH(SO4)1/2
5.4
0.5
−1.5809
ArgH(Citrate)1/2
3.3
0.5
−1.0850
ArgH(Acetate)
0.0
0.5
N/A
ArgHF
0.0
0.5
N/A
ArgHCl
0.0
0.5
−0.3357
ArgHBr
−7.7
0.5
0.0818
ArgHI
−22.6
0.5
0.2778
ArgHSCN
−24.2
0.5
0.5404
GdmCl
−6.6
0.5
N/A
NIH-PA Author Manuscript
NIH-PA Author Manuscript
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.
NIH-PA Author Manuscript
NIH-PA Author Manuscript
0.48
(Acetate)
0.48
0.48
(H2PO4)
0.48
0.48
(Citrate)1/2
Cl
0.49
(SO4)1/2
SCN
C (mol/L)
Arg Salt
−1.25
−1.80
−0.90
−2.00
−3.20
−0.90
−3.24
−2.13
−1.40
−4.63
−6.58
−0.99
B22 +−
B22 ++
0.40
0.45
−0.2
−1.60
−0.35
−4.89
B22 −−
−2.80
−2.35
−4.50
−8.80
−8.96
−20.73
Σij
McMillan-Mayer Second Virial Coefficient, B22 (L/mol), values for ion pairs in aqueous arginine salt solutions.
NIH-PA Author Manuscript
Table II
Schneider et al.
Page 24
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.
NIH-PA Author Manuscript
NIH-PA Author Manuscript
0.48
0.48
0.48
Cl
40.6
0.48
(H2PO4)
(Acetate)
SCN
35.8
0.48
(Citrate)1/2
38.2
49.4
29.3
42.7
0.49
(SO4)1/2
Cation-Cation
C (mol/L)
Arg Salt
4.0
6.3
30.6
30.1
31.7
51.3
Cation-Anion
0
0
0
11.7
0.4
0
Anion-Anion
0
0
8.3
19.5
28.2
67.3
Anion-Arg-Anion
Number of hydrogen bonds between different ions in aqueous arginine salt solutions.
NIH-PA Author Manuscript
Table III
Schneider et al.
Page 25
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.
NIH-PA Author Manuscript
NIH-PA Author Manuscript
0.5
(Acetate)
−4.5
−3
−7.7±1.5
−5.7±0.4
0.2±1.2
−2.6±0.3
4
−2.5
0
−4.5
−5.2±1.7
N/A
ΓMD
ΓVPO
4
−1
1
−2
−4
−4
Γ+
4
26.5
18.9
19.7
−1
−4
10
16.1
15.3
Protein-Arg h-bonds
−4
−3
−3
Γ−
The error bars on the preferential coefficient values are on the order of ±1
0.48
0.2
(H2PO4)
SCN
0.5
(Citrate)1/2
0.5
0.5
(SO4)1/2
Cl
C (mol/L)
Arg Salt
Theoretical preferential interaction coefficient values for α-Chymotrypsinogen A in aqueous arginine salt solutions.
NIH-PA Author Manuscript
Table IV
Schneider et al.
Page 26
J Phys Chem B. Author manuscript; available in PMC 2012 June 9.